AI-driven workplace decisions face growing legal scrutiny

A federal judge in Oakland declined to block Meta from laying off workers who say the company used artificial intelligence to help pick people for dismissal, keeping the dispute alive as one of the first major tests of AI in workplace decisions.

The lawsuit was filed by 26 current and former Meta employees in the U.S. District Court in Oakland, California. They allege that AI systems helped select workers for layoffs in a round that cut about 8,000 Meta employees, roughly 5% of the company’s workforce. The plaintiffs say the system disproportionately affected workers on medical, parental, pregnancy, family or disability leave, raising the possibility that the software scored employees while they were away on legally protected leave.

That claim goes to the heart of the legal fight: not only whether Meta’s selection process was fair, but whether the workers can show how an automated system reached its conclusions. In a traditional layoff process, employees may be able to point to a manager’s notes, performance reviews or direct statements. With algorithmic management, the key evidence can sit inside software logs, model outputs and vendor contracts that workers may never see. The case is forcing courts to confront whether an employer must keep enough records to explain an AI-assisted decision and whether employees can challenge a black-box system without access to its internal mechanics.

The broader stakes extend beyond Meta’s Menlo Park headquarters. Employers across the country are leaning on AI tools for screening applicants, monitoring productivity, ranking performance and flagging workers for termination because the systems are fast, scalable and often cheaper than human review. But the same traits make them difficult to audit. Workers want to know whether a person or a machine made the final call. Employers want to preserve flexibility and protect proprietary systems. That tension is now landing in labor and discrimination law, where the burden of proof may turn on evidence that is hidden from the people most affected.

The case also arrives as lawmakers and regulators debate whether companies should be required to explain automated decisions, retain records and provide appeal processes. If judges demand more documentation, employers may need to preserve logs and outputs for every AI-assisted layoff decision. If they do not, workers may be left trying to prove discrimination from data they cannot access, even as software takes a larger role in who stays and who goes.